For planning and food safety, farmers and economists depend on precise crop production projections. Models based on remote sensing have proven to be successful in predicting yields on a wide scale, but there is still a significant obstacle to their transferability across diverse geographic areas. In this work, we used 5 years of data (2019–2023) to predict corn and soybean yields in Indiana, USA, using a semi-physical modelling approach. The method described here utilizes photosynthetically active radiation (PAR), light use efficiency (LUE), and fraction of absorbed photosynthetically active radiation (FPAR) in calculating net primary productivity (NPP). Water stress, temperature stress, vegetative growth factor, and maximum light efficiency are used to compute the LUE. For corn and soybean, the model’s coefficient of determination (R2) was 0.76 and 0.56, respectively, with root mean square errors (RMSE) of 0.82 and 1.42 tons/ha. The results show the remote sensing model provided the yield for a region and made the model for enhancing the agricultural decision-making and resource management. The further improvement of this model parameter and high-resolution data can make the model more predictable for a regional yield, and we can use it for real-time crop monitoring.